The k   -orbit theory and polycirculant conjecture

Aleksandr Golubchik Osnabrueck, Germany,

15.03.2005

Abstract
The paper contains a proof of the conjecture of M. Klin and D. Maru s ˘   i c ˘   that an automorphism group of a transitive graph contains a permutation, decomposed in cycles of the same length. The proof is based on the k   -orbit theory developed by author. Key words: k   -orbits, partitions, permutations, symmetry, groups

1 Introduction

A permutation group G ( V )   is called 2   -closed if it is an automorphism group of a graph, another if it is an automorphism group of an ordered partition Q 2   of a set V 2   . If G < G   and G Q 2 = G Q 2 = Q 2   , then one says that G   is a 2   -closure for G   . A permutation is called regular if it is decomposed in cycles of the same length.
In [1P. Cameron described the conjecture of M. Klin and D. Maru s ˘   i c ˘   (Polycirculant conjecture) that was announced on the British Combinatorial Conference (Problem BCC15.12 (DM282)) in July 1995:
Theorem 1 Every 2   -closed transitive group contains a regular element.
Below we give a proof of this statement, based on the k   -orbit theory developed by author.
The first attempt to describe this theory was launched in [3.

2 k   -Orbit theory

Let G ( V )   be a permutation group of a degree n   and V ( k )   ( k n   ) be a non-diagonal part of V k   (i.e. all k   values of coordinates of a k   -tuple α k V ( k )   are different), then the action of G   on V ( k )   forms a partition G V ( k )   of V ( k )   on G   -invariant classes. This partition is called a system of k   -orbits of G   and we denote it as O r b k ( G )   .
The k   -orbit theory studies symmetry properties of k   -orbits that cannot be obtained from permutation group theory, studying a permutation group G   as a permutation algebra and therefore being not able to see the inside structure of a n   -orbit of G   and relations between k   -orbits of subgroups of G   . Also the graph representation of permutation groups that considers an action of a permutation group on V 2   is not able to see all relations in O r b k ( G )   for 2 < k n   .
The k   -orbit theory gives a new view on abstract finite groups and finite permutation groups and shows a way to simple solutions of some problems which either are not solved or have a complicated solution. These problems are: the polynomial solution of graphs isomorphism problem; the problem of a full invariant of finite groups, admitting a primitive (isomorphic) permutation representation; a simple proof of the Feit-Thompson theorem: the solvability of finite groups of odd order (that lies in the foundation of the Classification of finite simple groups [6); The simple proof of the Fein-Kantor-Schacher theorem: any transitive permutation group contains a fixed-point-free prime-power element [5; and the problem that we consider in this paper.
The specificity of the n   -orbit representation of a group G   is a possibility to do a group visible. In order to go to this visibility one makes a partition of a matrix of a n   -orbit on cells of k   -orbits of subgroups of an investigated permutation group G   and studies symmetry properties of cells and the whole partition. Below we consider some facts from the k   -orbit theory which are necessary for the proof of theorem  1 .
The k   -orbit theory considers different actions of permutations on k   -sets and symmetries that follows from those actions. Let α n = v 1 v n   be a n   -tuple, then the left action of a permutation g   on α n   is defined as g α n = g v 1 g v n   and the right action of g   on α n   is given as α n g = v g 1 v g n   (here V = [ 1 , n ]   ). From that we obtain the left and right action of a permutation on k   -tuples and k   -sets.
The left action of a permutation is (on definition) an isomorphism ( ( 123 ) { 123 , 132 } = { 312 , 321 }   ), the right action is not an isomorphism ( { 123 , 132 } ( 123 ) = { 312 , 213 }   ). If the right action is an isomorphism, then it is an indicium for existing of a subgroup with non-trivial normalizer, whose n   -orbit contains (right-)isomorphic k   -subsets connected with considered right action.
Let X n   be a n   -orbit of G   . One of n   -tuples from X n   we choose as an initial n   -tuple and consider all permutations from G   relatively to this n   -tuple. If α n   is an initial n   -tuple, then X n = G α n   . The specificity of an initial n   -tuple is the equality of number and order values of its coordinates (that we take from the same ordered set V   ). We call the initial n   -tuple as n   -space and its ordered subsets as subspaces. A k   -orbit X k   of G   is a projection of X n   on some subspace I k X k   . We write this projection as X k = p ^ ( I k ) X n   . So we consider subspaces as ordered sets of coordinates and k   -orbits as non-ordered sets of k   -tuples. The different orderings of k   -tuples (lines) in a matrix of a k   -orbit shows different symmetry properties of k   -orbit, related with corresponding properties of the investigated permutation group.
We say that two k   -sets X k   and X k   are G   -isomorphic if they are isomorphic and connected with a permutation g G   : X k = g X k   . If we study a group G   and X k , X k   are isomorphic, but not G   -isomorphic, then we say that they are S n   -isomorphic (where S n   is the symmetric group).
We define the left action G Y n   of a group G   on a n   -subset Y n   as G Y n { g Y n : g G }   . The left action A X n   of a subgroup A < G   on a n   -orbit X n   of G   is given as A X n { A α n : α n X n }   .
Correspondingly the right action Y n G { Y n g : g G }   and X n A { α n A : α n X n }   . The left action of G   on a k   -set Y k   and A   on a k   -orbit X k   of G   for k < n   is the same as for k = n   , because p ^ ( I k ) g α n = g p ^ ( I k ) α n   . By the right action the equality p ^ ( I k ) ( α n g ) = ( p ^ ( I k ) α n ) g   has senseonly if α n   is the initial n   -tuple. Hence the right action on k   -sets has not a direct reduction from the right action on n   sets. If X n   is a n   -orbit of G   , A < G   and Y n   is a n   -orbit of A   , then L n = G Y n = X n A   and R n = Y n G = A X n   are partitions of X n   on n   -orbits of left and right cosets of A   in G   , at that n   -orbits of left cosets are n   -orbits of conjugate with A   subgroups of G   and n   -orbits of right cosets are n   -orbits of A   . Projections of X n   , L n   and R n   on a subspace I k   gives a covering L k   and a partition R k   of the k   -orbit X k   on k   -orbits of left and right cosets of A   in G   . If Y k = p ^ ( I k ) Y n   , then L k = G Y k   and R k = A X k   . Below under action of permutations we understand the left action.
Let k   be a divisor of | G |   . We call k   as an automorphic number if there exists a k   -element suborbit of G   (a k   -element orbit of a subgroup of G   ).
Let C o ( α k )   be a set of coordinates of a k   -tuple α k   , then for a k   -set X k   we define C o ( X k )   as { C o ( α k ) : α k X k }   .
If C o ( α k )   is a suborbit of G   , then we say that α k   is an automorphic k   -tuple and C o ( α k )   is an automorphic subset of V   . Let X k   be a k   -set we say that X k   is an automorphic k   -set, if A u t ( X k )   acts on X k   transitive. Let X k   be a k   -orbit (an automorphic k   -set) and α k X k   be an automorphic k   -tuple, then a k   -orbit X k   we call as right-automorphic k   -orbit or k   -rorbit.
Let U   be a set and Q   be a set of subsets of U   , then Q ( W Q ) W   . Under Q   we understand a set of unions of intersected on U   classes of Q   . So Q   is a partition of Q   (possibly trivial). A union and intersection of partitions P   and R   of U   we write as P R   and P R   . Let P   be a subpartition of P   , then we write P P   .
Let α k X k   and Y k   be a maximal subset of X k   with C o ( Y k ) = C o ( α k )   , then we call Y k   as a k   -block of X k   .
Let C o ( X k ) = V   and C o ( X k )   be a non-trivial partition of V   , then we say that X k   is incoherent. If C o ( X k )   is a non-trivial covering of V   , then we say that X k   is coherent. Any coherent k   -orbit consists of intersected on V   k   -blocks. Classes of a partition of an incoherent k   -orbit are coherent k   -suborbits or k   -blocks. A coherent k   -orbit X k   is defined on a set V = C o ( X k )   . Its coherent or incoherent k   -suborbit Y k   is defined on a set U = C o ( Y k ) V   .
In order to show this we say that Y k   is U   -coherent ( U   -incoherent). If a ( V   -) coherent k   -orbit contains no U   -coherent and no U   -incoherent k   -suborbit for any U V   , then we call it as elementary coherent.
Let Y k   be a k   -suborbit of G   . The maximal transitive on Y k   subgroup of G   we call a stabilizer of Y k   in G   and write it as S t a b G ( Y k )   or simply S t a b ( Y k )   .
Let X k   be a k   -orbit of G   and Y k X k   be a k   -block, then evidently L k = G Y k   is a partition of X k   and Y k   is a k   -orbit of S t a b ( Y k ) = S t a b ( C o ( Y k ) )   .
Let a group G   be imprimitive, then G   contains a non-trivial ( 1 < k < n   ) incoherent k   -orbit.
If G   is primitive, then any (non-trivial) k   -orbit of G   is coherent. From here follows that it is convenient to consider primitive Abelian groups as trivial imprimitive or trivial primitive, because, as distinct from non-Abelian primitive groups, such a group G   contains no non-trivial suborbit U V   that forms a covering G U   of V   . So further under a primitive group we understand a non-Abelian primitive group. Let a group G   be imprimitive and Q   be a G   -invariant partition of V   ( G Q = Q   ), then we say that classes of Q   are imprimitivity blocks.
We say that a transitive group G ( V )   is a minimal degree group or md-group if it cannot be represented isomorphically on a partition Q   of V   , otherwise we say that G   is a non-minimal degree group or nmd-group. It follows that primitive groups are md-groups. Let F   be a finite group, A < F   and a representation F ( F / A )   be a md-group isomorphic to F   , then we say that A   is a md-stabilizer of F   . So A   is a maximal by including subgroup of F   containing no normal subgroup of F   .
Let α k = v 1 v k   and α l = v 1 v l   , then a ( k + l )   -tuple α ( k + l ) = v 1 v k v 1 v l   we call a concatenation of α k   and α l   and write this as α ( k + l ) = α k α l   .
Proposition 2 Let α k   be an automorphic k   -tuple, G = S t a b ( C o ( α k ) )   , H = S t a b ( α k )   and X k = G α k   , then H   is a normal subgroup of G   and a factor group G / H   is isomorphic to A u t ( X k )   .
Proof: If H   is not trivial, then G   is an intransitive group. Let X n   be a n   -orbit of G   , then there exists only one partition P n   of X n   on | G | / | H |   classes with projection p ^ ( α k ) P n = X k   .
Hence X n H = H X n = P n   . The group A u t ( X k )   is isomorphic to the factor group G / H   , because | A u t ( X k ) | = | G / H |   and it acts on cosets of H   in G   transitive.  
Proposition 3 Let X k O r b k ( G )   , Y k X k   be a k   -suborbit of G   , H = S t a b ( Y k )   , L k = G Y k   and R k = H X k   . Let L k = R k   , then H   is a normal subgroup of G   .
Proof: There exists only one partition P n   of X n   on | G | / | H |   classes so that p ^ ( I k ) P n = L k   .
Hence X n H = H X n = P n   .  
Proposition 4 Let G ( V )   be an imprimitive group, Q   be a partition of V   on imprimitivity blocks and Q   be ordered set Q   , then H = S t a b ( Q )   is a normal subgroup of G   .
Proof: Subgroups { S t a b ( U ) < G : U Q }   form a class of conjugate subgroups and H = ( U Q ) S t a b ( U )   . If G ( V )   is a nmd-group and G ( Q )   is an isomorphism, then H   is trivial.  
Proposition 5 Let X k O r b k ( G )   , Y k , Z k X k   be k   -suborbits of G   , L k = G Y k   and L k = G Z k   , then L k 1 = L k L k   and L k 2 = L k L k   are also partitions of X k   on isomorphic k   -suborbits of G   . If Y k Z k = T k   be not trivial and U k L k 2   contains Y k   and Z k   , then L k 1 = G T k   , L k 2 = G U k   , S t a b ( T k ) = S t a b ( Y k ) S t a b ( Z k )   and S t a b ( U k ) = g r ( S t a b ( Y k ) , S t a b ( Z k ) )   .
Proof: Let k = n   , then equalities follows from corresponding properties of sets of left cosets of subgroups of G   . For k < n   equalities are projections of corresponding equalities for k = n   .  
Lemma 6 Let for every k   -suborbit Y k   of a k   -orbit X k   a set L k = A u t ( X k ) Y k   be a partition of X k   , then | A u t ( X k ) | = | X k |   .
Proof: Let G = A u t ( X k )   , M   be a set of partitions of X k   on isomorphic k   -suborbits of G   and L k , L k M   , then L k 1 = L k L k   and L k 2 = L k L k   are partitions from M   . Let L k M   be a partition of X k   on k   -blocks, then there exists a partition L k M   so that L k 2 = L k L k   is not trivial, i.e. | L k 2 | < m i n ( | L k | , | B k | )   . By repeating we obtain that there exist partitions L k , L k M   so that L k 2 = L k L k = { X k }   .
For each partition L k   in the union process there exists a subgroup A < G   so that a class Y k L k   is a k   -orbit of A   and | A | = | Y k |   , because we can begin from a subgroup that is isomorphic to the automorphism group of a k   -block, and with a subgroup of a prime order that permutes k   -blocks. Thus G   contains a subgroup H   of order | X k |   that acts transitive on X k   .
Let g   be any permutation from G   and R k = g r ( g ) X k   , then ( g r ( g ) B ) X k = R k   , because any class Z k   of R k   generates a partition G Z k M   . It follows that the action of G   and B   on X k   are isomorphic and hence G = B   .  
Proposition 7 Let X k   be an incoherent automorphic k   -set, then | A u t ( X k ) | / | X k | > 1   .
Proof: If X k   is incoherent, then A u t ( X k )   has evidently a permutation that fixes a k   -tuple from X k   .  
Lemma 8 Let X k   be an elementary coherent k   -orbit, then | A u t ( X k ) | = | X k |   .
Proof: Let Y k X k   be a k   -block and Z k X k   be a k   -suborbit of A u t ( X k )   that has non-trivial intersection with Y k   . Let A = g r ( S t a b ( Y k ) , S t a b ( Z k ) )   , then k   -orbit of A   in X k   is X k   , because the elementary coherent k   -orbit X k   is the unique super- k   -suborbit for any k   -block.
Hence A u t ( X k ) Z k   is a partition of X k   . Then the statement follows from lemma  6 .  
Lemma 9 Let G   be a transitive group, a n   -tuple α n = α k α k α m   , 2 k + m = n   , α k   and α k   be automorphic and G   -isomorphic, and C o ( α m )   contains no k   -tuple G   -isomorphic to α k   , then
  • the ( 2 k )   -tuple α 2 k = α k α k   is automorphic,
  • the group A = S t a b ( C o ( α k ) ) S t a b ( C o ( α k ) )   is not trivial,
  • the group A   has a normalizer N   that acts imprimitive on C o ( α 2 k )   with imprimitivity blocks C o ( α k )   and C o ( α k )   .
Proof: Let X n O r b n ( G )   , X k = p ^ ( α k ) X n   , X k = p ^ ( α k ) X n   and X 2 k = p ^ ( α 2 k ) X n   . Let Y k , Y k X k = X k   be k   -blocks with coordinates C o ( Y k ) = C o ( α k )   and C o ( Y k ) = C o ( α k )   . Consider a maximal ( 2 k )   -subset Y 2 k X 2 k   , whose projection on α k   is Y k   and a maximal ( 2 k )   -subset Y 2 k X 2 k   , whose projection on α k   is Y k   . The ( 2 k )   -subset Y 2 k   can be partition on ( 2 k )   -orbits of S t a b ( C o ( α k ) )   and the ( 2 k )   -subset Y 2 k   can be partition on ( 2 k )   -orbits of S t a b ( C o ( α k ) )   . Take in opinion that α k   and α k   have on condition non-trivial intersection with any k   -tuple from X k \ Y k \ Y k   and that the intersection Y 2 k Y 2 k   contains α 2 k   , we obtain that Z 2 k = Y 2 k Y 2 k   is a ( 2 k )   -orbit of the non-trivial subgroup A   .
Let now T 2 k , T 2 k X 2 k   be maximal ( 2 k )   -subsets with projection Y k Y k   on α k   and α k   correspondingly, then U 2 k = T 2 k T 2 k   is a ( 2 k )   -orbit of the imprimitive on α 2 k   normalizer N   of A   with imprimitivity blocks C o ( α k )   and C o ( α k )   .   For the greater than two automorphic G   -isomorphic subsets we have more possibilities for permutations of subsets and therefore find more possible properties.
Lemma 10 Let α n = α k 1 α k l α m   , r = l k   , n = r + m   , all α k i   are automorphic and G   -isomorphic and no class of G C o ( α k 1 )   belongs to C o ( α m )   , then r   -tuple α r = α k 1 α k l   is automorphic, a subgroup A = ( i = 1 , l ) S t a b ( C o ( α k i ) )   can be trivial and has a normalizer N   that acts imprimitive on α r   with imprimitivity blocks C o ( α k i )   .
Proof: Let X k = p ^ ( α k i ) X n   and X r = p ^ ( α r ) X n   . If g G   maps α k 1   on α k 2   , then α k 2   can be mapped on any k   -tuple with coordinates C o ( α k i )   . So in this case A   can be trivial, but N   is not trivial, because subsets C o ( α k i )   are non-intersected in pairs and, on condition, they are intersected with any k   -tuple from X k \ ( i = 1 , l ) Y k i   , where Y k i   are k   -blocks with C o ( Y k i ) = C o ( α k i )   .
 
Theorem 11 Let G ( V )   be a primitive group and k   be a maximal automorphic divisor of n   , then there exists a partition Q   of V   on automorphic G   -isomorphic k   -subsets or on automorphic S n   -isomorphic k   -subsets.
Proof: Let the statement be not correct, then there exists a subset U   of V   that can be partition on automorphic G   -isomorphic k   -tuples { α k , } = P   ( where 1 < | P | < n / k   ), and the subset W = V \ U   contains no automorphic k   -tuple G   -isomorphic to α k   , then from lemma  10 immediately follows that U   is an automorphic subset. If | U |   divides n   , then k   is not the maximal divisor of n   . Contradiction. So | U |   does not divide n   and Q = G U   is a covering of V   .
Let Q = G C o ( α k )   , then Q   and Q   are G   -invariant and therefore an intersection of any two classes of Q   contains whole number of k   -tuples from Q   . Hence W   must contain some class from Q   . Contradiction.  
Lemma 12 A nmd-group is 2   -closed.
Proof: Let G ( V )   be not 2   -closed a nmd-group and A ( V )   be a 2   -closure of G ( V )   , then A ( V )   and G ( V )   have the same set of imprimitivity blocks Q   and hence A ( V )   is a md-group.
Let G ( Q )   be a md-reduction of G ( V )   and A ( Q )   be a md-reduction of A ( V )   . Since G ( V )   is not 2   -closed, then G ( Q )   is not 2   -closed too and hence A ( Q ) G ( Q )   . It follows that A ( Q )   is isomorphic to A ( V )   , i.e. A ( V )   is a nmd-group. Contradiction.  
Lemma 13 Let Q   be a set of imprimitivity blocks of a nmd-group G ( V )   and C o ( I k ) Q   , then S t a b ( I k )   is trivial.
Proof: Let Q = { C o ( I k i ) : i [ 1 , | Q | ] }   , X n O r b n ( G )   , X k = p ^ ( I k i ) X n   and X 2 k i j = p ^ ( I k i I k j ) X n   , then from lemma  12 it follows that G = ( i , j [ 1 , | Q | ] ) A u t ( X 2 k i j )   . Let | X n | / | X k | > 1   ,then there exists a partition L 2 k i j   of X 2 k i j   so that p ^ ( I k i ) L 2 k i j = p ^ ( I k j ) L 2 k i j = { X k }   . It follows that A u t ( X 2 k i j )   and hence G   contain a subgroup H   of order | X k |   . Then there exists only one partition P n   of X n   on | X n | / | X k |   automorphic classes so that p ^ ( I k i ) P n = { X k }   , hence H   is a normal subgroup of G   . Let a representation G ( Q )   be a md-group, then S t a b ( C o ( I k ) )   is a md-stabilizer of G   . But S t a b ( C o ( I k ) )   has non-trivial intersection with H   and hence it cannot be a md-stabilizer of G   . Contradiction.  
Lemma 14 Let G ( V )   be a nmd-group, then there exists an isomorphic representation G ( Q )   of G ( V )   on imprimitivity blocks of a prime power.
Proof: Let | V | / | Q | = k   be not prime and p   be a prime divisor of k   . Let C o ( I k ) Q   , then I k   contains an automorphic p   -tuple I p   , so that Q = G C o ( I p ) G C o ( I k ) = Q   . Hence G   has an imprimitive representation on classes of Q   . If G ( Q )   is a homomorphism, then G ( Q )   is a homomorphism too. Contradiction. Thus G ( Q )   is an isomorphism.   This statement can be generalized:
Lemma 15 Let Q   be a partition of V   and G ( Q )   be a homomorphism of G ( V )   , then there exists a partition Q Q   on subsets of V   of a prime power so that G ( Q )   is also a homomorphism, at that if G ( Q )   is an isomorphism, then G ( Q )   is an isomorphism too.

3 The proof of theorem  1 

3.1 Imprimitive nmd-groups

From lemma  14 it follows that any nmd-group contains a regular element.

3.2 Primitive groups

Let G   be a primitive group. If G   contains a nmd-subgroup, then it contains a regular element.
Let G   contains no nmd-subgroup, but an imprimitive md-subgroup A   .
Lemma 16 Let A   be a maximal imprimitive md-subgroup of G   , then G   is 2   -closed if and only if A   is 2   -closed.
Proof: Let Q   be a set of imprimitivity blocks of A   , then A = S t a b ( Q )   and all classes of G Q   (partitions of V   ) are different. Let X n O r b n ( G )   , Y n O r b n ( A )   and L n = G Y n   . Let A   be not 2   -closed, then any class Z n L n   is not 2   -closed and hence X n   is not 2   -closed. Let X n   be not 2   -closed and X n   be a 2   -closure of X n   , then X n   contains an imprimitive on Q   n   -suborbit Y n   that is a 2   -closure of Y n   .   Thus a consideration of a primitive group G   can be reduced to a consideration of its imprimitive md-subgroup. So we have to consider a primitive group that contains no imprimitive subgroup.
Lemma 17 Let G   contains no imprimitive subgroup, then it is not 2   -closed.
Proof: If G   contains no imprimitive subgroup, then (lemma  10 and theorem  11 ) there exists a partition Q   of V   on automorphic S n   -isomorphic k   -subsets and no partition of V   on automorphic G   -isomorphic k   -subsets. Let X n O r b n ( G )   , Q = { C o ( α k i ) : i [ 1 , | Q | ] ) }   , X k i = p ^ ( α k i ) X n   and L k = { X k i : i [ 1 , | Q | ] ) }   , then any two k   -blocks from a k   -orbit X k i L k   are intersected on V   . Hence a non-trivial normalizer N = S t a b S n L k   of G   is a 2   -closure of G   .   So we can reduce the consideration of theorem  1 to imprimitive md-groups.

3.3 Imprimitive md-groups

Let G ( V )   be an imprimitive md-group and Q   be a partition of V   on imprimitivity blocks, then a group H = S t a b ( Q )   is not trivial. Let Y n O r b n ( H )   , Q = { C o ( I k i ) : i [ 1 , | Q | ] }   , Y 2 k i j = p ^ ( I k i I k j ) Y n   and G   be 2   -closed, then H   is also 2   -closed and hence H = ( i , j [ 1 , | Q | ] ) A u t ( Y 2 k i j ; V )   , where A u t ( Y 2 k i j ; V )   is an extension of A u t ( Y 2 k i j )   on V   . Accordingly to lemma  15 we can assume that | V | / | Q | = p   is prime, then A u t ( Y 2 k i j ; V )   (for any i , j [ 1 , | Q | ]   ) is a direct product and contains a regular element of order p   . Hence H   contains a regular element of order p   .
Conclusion The main inference from the investigation of k   -orbits symmetry properties is that a finite group F   is not closed by its algebraic properties, because the group algebra F   is equivalent to the action of F   on F   , but this algebra generates also the action of F   on F k   . Properties of that action not always can be interpreted with group algebra or with traditional permutation group theory. Namely such properties are the subject of investigation of the k   -orbit theory. Some investigations related with an application of k   -orbit representation to problems announced above are described in [2- [4The k   -orbit theory gives a new view on finite many-dimensional symmetries and can bring new ideas and applications to other sciences studying symmetrical objects.
References

  1. Peter J. Cameron, Elusive groups and the polycirculant conjecture, Queen Mary, University of London, Barcelona, February 2001
  2. Aleksandr Golubchik, The polynomial algorithm for graphs' isomorphism testing, , February 2002. ☻ open access ✓
  3. Aleksandr Golubchik, On the Polycirculant conjecture, , April 2002. ☻ open access ✓
  4. Aleksandr Golubchik, On the nature of finite groups, , September 2004. ☻ open access ✓
  5. Aleksandr Golubchik, The k   -orbit theory and Fein-Kantor-Schacher Theorem, , February 2005. ☻ open access ✓
  6. Daniel Gorenstein, Finite simple groups, Plenum Press, New York and London, 1982